Methods and apparatus for training a Neural Network to recover a codeword of a Forward Error Correction (FEC) code are provided. Trainable parameters of the Neural Network are optimised to minimise a loss function. The loss function is calculated by representing an estimated value of the message bit output from the Neural Network as a probability of the value of the bit in a predetermined real number domain and multiplying the representation of the estimated value of the message bit by a representation of a target value of the message bit. Training a neural network may be implemented via a loss function.

Certain aspects of the present disclosure relate to techniques and apparatus for improving decoding latency and performance of Polar codes. An exemplary method generally includes generating a codeword by encoding information bits, using a multi-dimensional interpretation of a polar code of length N, determining, based on one or more criteria, a plurality of locations within the codeword to insert error correction codes generating the error correction codes based on corresponding portions of the information bits, inserting the error correction codes at the determined plurality of locations, and transmitting the codeword. Other aspects, embodiments, and features are also claimed and described.

A memory device can include a memory array, a processor coupled to the memory array, and a decoding apparatus. The decoding apparatus is configured to perform parallel decoding of codewords. Each of the codewords has a plurality of data blocks, and each data block having a number of data bits. The decoding apparatus is configured to decode, in parallel, a first codeword with one or more other codewords to determine error information associated with each codeword. For errors in a common data block shared between two codewords being decoded in parallel, the error information includes a data block identifier and associated error bit patterns. Further, the decoding apparatus is configured to update the codewords based on the error information.

There is provided a method of recursive sequential list decoding of a codeword of a polar code comprising: obtaining an ordered sequence of constituent codes usable for the sequential decoding of the polar code, representable by a layered graph; generating a first candidate codeword (CCW) of a first constituent code, the first CCW being computed from an input model informative of a CCW of a second constituent code, the first constituent code and second constituent code being children of a third constituent code; using the first CCW and the second CCW to compute, by the decoder, a CCW of the third constituent code; using the CCW of the third constituent code to compute a group of symbol likelihoods indicating probabilities of symbols of a fourth (higher-layer) constituent code having been transmitted with a particular symbol value, and using the group of symbol likelihoods to decode the fourth constituent code.

There is provided mechanisms for decoding an encoded sequence into a decoded sequence. A method is performed by an information decoder. The method comprises obtaining the encoded sequence. The encoded sequence has been encoded using a polar code. The method comprises successively decoding the encoded sequence into the decoded sequence. The decoding is performed for a given list size, LS, where LS>1, defining how many candidate decoded sequences in total the thus far decoded sequence is allowed to branch into during the decoding. The encoded sequence is decoded, until its first branching, by at least as many processing units in parallel as a factor, f, of the given list size. The factor is at least half the given list size, f≥LS/2.

The disclosed structures and methods are directed to decoders and to methods for decoding codes, for example, polar codes. The decoder comprises: a codeword node decoding pipeline having three logic units, and configured to, for each encoded codeword node: based on a received instruction sequence, adjust the three logic units for decoding of each encoded codeword node, and decode a set of logarithmic likelihood ratios (LLRs) corresponding to the encoded codeword node to generate decoded bits. The decoder also has an output storage configured to store the decoded bits corresponding to each encoded codeword node, and generate a decoded codeword based on the decoded bits. The decoding method comprises adjusting the codeword node decoding pipeline to each encoded codeword node based on codeword node length and a codeword node type, as well as a bit index of the encoded codeword node.

Methods, systems, and devices for wireless communications employing multi-level coding with set partitioning on transmitting side and multi-level sequential decoding on the receiving side for latency minimization are described. In some systems, a transmitting device may transmit a code block group (CBG) to a receiving device including a first set of code blocks associated with a first decoding level and a second set of code blocks associated with a second decoding level. The receiving device may unsuccessfully decode one or more code blocks of the first set or the second set of code blocks and transmit a feedback message to the transmitting device. The transmitting device may determine that the data to be communicated via the CBG is latency-sensitive data and, as such, determine to retransmit both the first set and the second set of code blocks to the receiving device in response to receiving the feedback message.

The disclosed structures and methods are directed to decoders and to methods for decoding codes, for example, polar codes. The decoder comprises: a codeword node decoding pipeline having three logic units, and configured to, for each encoded codeword node: based on a received instruction sequence, adjust the three logic units for decoding of each encoded codeword node, and decode a set of logarithmic likelihood ratios (LLRs) corresponding to the encoded codeword node to generate decoded bits. The decoder also has an output storage configured to store the decoded bits corresponding to each encoded codeword node, and generate a decoded codeword based on the decoded bits. The decoding method comprises adjusting the codeword node decoding pipeline to each encoded codeword node based on codeword node length and a codeword node type, as well as a bit index of the encoded codeword node.

Methods, systems, and devices for wireless communications employing multi-level coding with set partitioning on transmitting side and multi-level sequential decoding on the receiving side for latency minimization are described. In some systems, a transmitting device may transmit a code block group (CBG) to a receiving device including a first set of code blocks associated with a first decoding level and a second set of code blocks associated with a second decoding level. The receiving device may unsuccessfully decode one or more code blocks of the first set or the second set of code blocks and transmit a feedback message to the transmitting device. The transmitting device may determine that the data to be communicated via the CBG is latency-sensitive data and, as such, determine to retransmit both the first set and the second set of code blocks to the receiving device in response to receiving the feedback message.

Systems and methods are disclosed for performing polar encoding of a number of information bits for transmission in a wireless communication system in a manner that is optimized for a specific code length. In some embodiments, a method of operation of a transmit node in a wireless communication system comprises performing polar encoding of a set of K information bits to thereby generate a set of polar-encoded information bits where the K information bits are mapped to the first K information bit locations specified in an information sequence S.sub.N which is a ranked sequence of N information bit locations among a plurality of input bits for the polar encoding where N is equivalent to a code length, a size of the information sequence S.sub.N is greater than or equal to K, and the information sequence SN is optimized for a specific value of the code length N.